Re: [Sidrops] WGLC for draft-ietf-sidrops-ov-egress-00.txt - ENDS 11/25/2019 (November 25 2019)

["Sriram, Kotikalapudi (Fed)" <[e-mail]>]

>that seems fair. (provided that this was what Sriram meant I guess)

My problem was with these two sentences:

#1 
"Configurations may have complex policy where the final announced
   origin AS may not be easily predicted before all policies have been
   run."

Is it not the same as simply saying: 
Configurations may have complex policy where the final announced
   origin AS is determined only after all policies have been run.

Why not state that and keep it simple?

#2 
"Therefore it SHOULD be possible to specify an origin validation
   policy which MUST BE run after such non-deterministic policies."

What does it mean to say "specify origin validation policy"?
RFC 6811 has already specified the origin validation procedure (policy?).
The use of the term "non-deterministic policies" is ill advised.
The operator knows their complex policies and can determine
the origin AS that would result after the policies are applied.
So the sentence can be simply stated as:
Therefore, origin validation MUST BE performed after the policies are run.

I hope this clarifies what I was trying to say.

Sriram




________________________________________
From: Chris Morrow <[e-mail]>
Sent: Sunday, December 1, 2019 4:31 PM
To: Randy Bush
Cc: Chris Morrow; Sriram, Kotikalapudi (Fed); [e-mail]; [e-mail]; [e-mail]; Jeffrey Haas
Subject: Re: [Sidrops] WGLC for draft-ietf-sidrops-ov-egress-00.txt - ENDS 11/25/2019 (November 25 2019)

On Sun, 01 Dec 2019 20:40:14 +0000,
Randy Bush <[e-mail]> wrote:
>
> Many documents already say to be sure to issue ROA(s) for the AS(s)
> which announce the prefix. This document does not change that in the
> slightest. If we start reiterating practice and restating the obvious
> we will have a document three or more times as large but with zero
> real added value.  Let us not do that to this simple document.
>

that seems fair. (provided that this was what Sriram meant I guess)

> randy, on a phone
>
> > On Dec 1, 2019, at 12:07, Chris Morrow <[e-mail]> wrote:
> >
> > On Wed, 27 Nov 2019 21:17:08 +0000,
> >>> Is there such a thing as non-deterministic policy?
> >>> What does it mean to say that the outcome of applying
> >>> one or a set of policies is that it results in
> >>> a random AS being the origin AS?
> >>
> >> non-derministic != random
> >>
> >> in this case it is meant to refer to policy results which depend on
> >> exogenous data
> >
> > The case probably is that at some point (with certain attributes set
> > by an upstream bgp speaker) OriginAS could be any of a set of values.
> >
> > So, ideally the operator would just create ROA with the set covered,
> > right? (really 1 ROA per origin)
> >
> > I think Sriram is getting to that, or is saying it in a way that
> > wasn't clear to me :)
> >
> > -chris